Analysis of hematologic toxicity profiles following CAR-T cell therapy: a retrospective cohort study

Abstract Background Chimeric antigen receptor (CAR) T-cell therapy has transformed the treatment of relapsed/refractory (R/R) B-cell malignancies, most notably large B-cell lymphoma. However, hematologic toxicity—recently formalized as immune effector cell-associated hematotoxicity (ICAHT), comprising neutropenia (N-ICAHT) and thrombocytopenia (T-ICAHT)—remains a major clinical challenge. We aimed to characterize the incidence and predictors of N-ICAHT and T-ICAHT in a real-world cohort using the ICAHT and T-ICAHT grading frameworks. Methods This single-center retrospective cohort study included 96 consecutive patients with R/R B-cell malignancies treated with CAR T-cell therapy between November 2022 and March 2025. Hematologic toxicity was graded according to CTCAE v4.0.3 and the ICAHT/T-ICAHT systems. Patients were stratified into low-risk (CAR-HEMATOTOX 0–1) and high-risk (≥ 2) groups. Candidate predictors of severe N-ICAHT and T-ICAHT were selected using LASSO regression and entered into multivariable logistic models. Results The median age was 56.5 years (IQR 46–66.5). The CAR-HEMATOTOX score was calculable in 86 patients, of whom 31 (36.0%) were high-risk (≥ 2). The overall response rate (ORR) was 78.3% at 1 month and 65.5% at 3 months. Cytokine release syndrome (CRS) occurred in 74.0% of patients (grade ≥ 3, 2.1%) and immune effector cell-associated neurotoxicity syndrome (ICANS) in 10.4% (grade ≥ 3, 5.2%). Grade 3–4 neutropenia peaked at day 4 (71.9%) and persisted in 34.7% at day 28. Severe early N-ICAHT (grade 3–4) was more frequent in high-risk than low-risk patients (16.1% vs. 3.6%, p = 0.013). In multivariable analysis, CAR-HEMATOTOX ≥ 2 was the only significant predictor of severe early N-ICAHT (adjusted OR 4.04, 95% CI 1.01–22.17), whereas bridging therapy predicted late N-ICAHT (adjusted OR 4.25, 1.64–10.98). For T-ICAHT, high-risk patients had higher rates of both early (74.2% vs. 25.5%, p < 0.001) and late (80.6% vs. 47.3%, p < 0.001) thrombocytopenia; CAR-HEMATOTOX ≥ 2 was the strongest adjusted predictor of early (adjusted OR 8.22, 2.73–22.29) and late (adjusted OR 4.20, 1.57–13.29) T-ICAHT. Notably, 9 of 10 patients with ICANS (90%) developed T-ICAHT. G-CSF (78.1%) and thrombopoietin receptor agonist (TPO-RA; 48.9%) support were not associated with increased grade ≥ 3 CRS or ICANS. Conclusion In this real-world cohort, the CAR-HEMATOTOX score predicted both N-ICAHT and T-ICAHT, with a stronger association for thrombocytopenia than neutropenia. The observed association between ICANS and T-ICAHT is hypothesis-generating and may reflect shared endothelial pathophysiology, although residual confounding cannot be excluded. Within the limits of this observational design, G-CSF and TPO-RA support were not associated with increased severe CRS or ICANS.